The government miscalculated the number of U.S. troops who may have been exposed to nerve gases when Iraqi weapons were destroyed during the first Gulf War, congressional investigators say.

The General Accounting Office (search) is expected to testify in a House hearing Monday that the Pentagon and CIA used a flawed computer model to estimate the fallout from the weapons. The models were created with inaccurate data, and the height of the plume resulting from the 1991 weapons explosions was underestimated, according to a memo sent to members of a House Government Reform subcommittee.

The Associated Press obtained a copy of the memo Friday.

The data indicate that initial reports that about 100,000 troops were exposed were wrong. The memo doesn't say whether more or fewer troops were likely to have been affected.

"We shouldn't be making policy by relying on that modeling," said Larry Halloran, a spokesman for the National Security, Emerging Threats and International Relations subcommittee (search).

The GAO finding is sure to reverberate among Gulf War veterans and their advocates, who have said the exposure to the weapons' gases was one cause of mysterious illnesses among many veterans. Advocates also will use the testimony to argue for health care and compensation for veterans within a certain radius of the weapons explosions.

Veterans of the 1991 Gulf War have suffered from illnesses they believe are linked to their service in Gulf. Among reported symptoms are chronic fatigue, diarrhea, migraines, dizziness, memory problems, loss of muscle control and loss of balance.

"We have to presume that the illnesses the veterans have are related to that exposure. We have to lean toward the veteran," said Steve Robinson, National Gulf War Resource Center (search) executive director.

The findings also could have implications for research on illnesses reported by Gulf War veterans. GAO concluded that epidemiological studies that relied on the computer model would be invalid.

U.S. troops destroyed ammunitions caches in Khamisiyah, Iraq, on March 4 and March 10, 1991. Some of those weapons contained sarin (search) and cyclosarin, two nerve gases.

The Pentagon has said more than 100,000 soldiers were exposed to low levels of nerve gas when the weapons were destroyed in Khamisiyah.

The computer model used was developed by a panel of experts convened by the Department of Defense's Institute for Defense Analyses (search). The panel used several computer models and combined them to make up for the lack of onsite measurements of chemical exposure and local weather data. But the GAO found the models had differed greatly on the size and path of the plume, the congressional memo said.

GAO found that the Pentagon did not have accurate weather data to model Khamisiyah and that made determining the speed and direction of the plume difficult, according to the subcommittee memo.

Monday's hearing, to be led by Rep. Christopher Shays, R-Conn., is designed to examine computer modeling's validity in tracking biological, chemical or radiological substances. The National Research Council will release a report on plume modeling, Halloran said.

"The lessons are that at least with the current state of science, (computer modeling) is particularly ineffective for any retrospective analysis," Halloran said. "You need data collection that will allow you to eliminate all kinds of variables so you can warn people where (the weapon's plume) is going."